US6122321A - Methods and apparatus for reducing the complexity of inverse quantization operations - Google Patents

Abstract

Methods and apparatus for implementing video decoders at a lower cost than known video decoders are described. The methods include data reduction techniques, simplified inverse quantization techniques, and dynamically varying prediction filter complexity as a function of whether luminance or chrominance data is being processed. In various embodiments, data representing portions of B frames which will not be displayed is identified and discarded, e.g., without performing a decoding operation thereon. Portions of I and P frames which will not be displayed are identified and decoded at a reduced resolution and/or using simplified inverse quantization techniques. The decoded I and P frame data is stored for use when making subsequent predictions if required. Another data reduction technique involves applying different amounts of downsampling to different image portions with no or little downsampling being performed on image portions located at or nearest to the image's center of interest as determined from information included in a received encoded bitstream. Simplified inverse quantization techniques of the present invention which are suitable for use when processing data from P or B frames involve performing a simplified non-MPEG compliant inverse quantization operation. The inverse quantization operation of the present invention is implemented by treating an additive factor, k, as having a value of zero for both predictively and intra-coded data and/or by eliminating the mismatch control operation normally performed as part of an MPEG compliant inverse quantization operation.

Description

FIELD OF THE INVENTION

The present invention relates to video decoders, and more specifically, to methods and apparatus for reducing the cost of video decoders, e.g., MPEG compatible video decoders.

BACKGROUND OF THE INVENTION

Various digital applications, such as digital video, involve the processing, storage, and transmission of relatively large amounts of digital data representing, e.g., one or more digital images.

In order to reduce the amount of digital data that must be stored and transmitted in conjunction with digital applications, various digital signal processing techniques have been developed. These techniques involve, e.g., the variable length encoding of data, quantization of data, transform coding of data, and the use of motion compensated prediction coding techniques.

One standard proposed for the coding of motion pictures, commonly referred to as the MPEG-2 standard, described in ISO/IEC 13818-2 (Nov. 9, 1994) Generic Coding of Moving Picture and Associated Audio Information: Video(hereinafter referred to as the "MPEG" reference), relies heavily on the use of DCT and motion compensated prediction coding techniques. An earlier version of MPEG, referred to as MPEG-1 also supports the use of the above discussed coding techniques. In this patent application, references to MPEG-2 compliant data streams and MPEG-2 compliant inverse quantization operations are intended to refer to data streams and inverse quantization operations that are implemented in accordance with the requirements set forth in the MPEG reference.

Another standard which also uses motion compensated prediction, referred to herein as the ATSC standard, is specifically intended for television applications. The ATSC standard is described in a document identified as ATSC A/53 titled "ATSC DIGITAL TELEVISION STANDARD" (1994).

In accordance with both the MPEG and ATSC standards images, e.g., frames, can be coded as intra-coded (I) frames, predictively coded (P) frames, or bi-directional coded (B) frames. I frames are encoded without the use of motion compensation. P frames are encoded using motion compensation and a reference to a single anchor frame. In the case of a P frame, the anchor frame is a preceding frame in the sequence of frames being decoded. B frames are encoded using a reference to two anchor frames, e.g., a preceding frame and a subsequent frame. Reference to the preceding frame is achieved using a forward motion vector while reference to the subsequent frame is achieved using a backward motion vector. In MPEG, I and P frames may be used as anchor frames for prediction purposes. B frames are not used as anchor frames.

A known full resolution video decoder 2, i.e., an MPEG-2 video decoder, is illustrated in FIG. 1. As illustrated, the known decoder 2 includes amemory 30, a syntax parser and variable length decoding (VLD)circuit 4,inverse quantization circuit 16, inverse discretecosine transform circuit 18,summer 22,switch 24, pair of motion compensatedprediction modules 6 and 7 and a select/average predictions circuit 28 coupled together as illustrated in FIG. 1. Thememory 30 includes both a codeddata buffer 32 which is used for storing encoded video data and areference frame store 34 used for storing decoded frames which may be used, e.g., as anchor frames and/or output to a display device.

For commercial reasons, particularly in the case of consumer products, it is desirable that video decoders be inexpensive to implement. The complexity of a decoder, in terms of the number of elements required to implement a decoder, is a factor which affects implementation costs. For this reason, an important issue in the design and implementation of video decoders is the minimization of complexity in terms of the amount of hardware, e.g., logic gates, required to implement a decoder.

A number of methods for reducing the complexity and/or cost of decoders have been developed. Examples of known method of reducing the cost of decoders include, e.g., the use of a preparser, the use of downsampled frames as prediction references, and the efficient implementation of the IDCT operation with downsampling. Video decoders which perform downsampling are sometimes referred to as "downsampling video decoders". Downsampling video decoders are discussed in U.S. Pat. No. 5,635,985 which is hereby expressly incorporated by reference.

FIG. 2 illustrates a knowndownsampling video decoder 10. Thedecoder 10 includes apreparser 12, a syntax parser and variable length decoding (VLD)circuit 14, aninverse quantization circuit 16, an inverse discrete cosine transform (IDCT)circuit 18, adownsampler 20,summer 22,switch 24,memory 30, a pair of motion compensatedprediction modules 25, 27 and a select/average predictions circuit 28. The motion compensatedprediction modules 25, 27 may include adrift reduction filter 26. Thememory 30 includes a codeddata buffer 32 and areference frame store 34. The various components of thedecoder 10 are coupled together as illustrated in FIG. 2.

In the knowndecoder 10, thepreparser 12 receives encoded video data and selectively discards portions of the received data prior to storage in the codeddata buffer 32. The encoded data from thebuffer 32 is supplied to the input of the syntax parser andVLD circuit 14. Thecircuit 14 provides motion data and other motion prediction information to the motion compensatedprediction modules 25, 27. In addition, it parses and variable length decodes the received data. Themodules 25, 27 each include a motion compensatedprediction filter 26. A data output of the syntax parser andVLD circuit 14 is coupled to an input of theinverse quantization circuit 16.

Theinverse quantization circuit 16 generates a series of DCT coefficients which are supplied to theIDCT circuit 18. From the received DCT coefficients, theIDCT circuit 18 generates a plurality of integer pixel values. In the case of intra-coded images, e.g., I frames, these values fully represent the image being decoded. In the case of inter-coded images, e.g., P and B frames, the output of theIDCT circuit 18 represents image (difference) data which is to be combined with additional image data to from a complete representation of the image or image portion being decoded. The additional image data, with which the output of the IDCT circuit is to be combined, is generated through the use of one or more received motion vectors and stored reference frames. The reference frames are obtained by theMCP modules 25, 27 from thereference frame store 34.

In order to reduce the amount of decoded video data that must be stored in thememory 30, thedownsampler 20 is used. In the case of inter-coded data, the downsampled video data output by thedownsampler 20 is stored, viaswitch 24, in thereference frame store 34.

Motion compensatedprediction modules 25 and 27 receive motion vector data from the syntax parser andVLD circuit 14 and downsampled anchor frames from thereference frame store 34. Using these inputs, they perform motion compensated prediction operations.

In the case of uni-directional motion compensation, the output of one ofmodules 25, 27 is selected by the select/average prediction circuit 28 and supplied to thesummer 22. In the case of bi-directional motion compensation the values output by themodules 26 and 27 are averaged by theaverage predictions circuit 28. The values generated by thecircuit 28 are supplied to the input of thesummer 22.

In the case of inter-coded video data, thesummer 22 is used to combine the output of thedownsampler 20, with the output of the select/average predictions circuit 28. The resulting data which represents a decoded inter-coded video frame is stored, viaswitch 24, in thereference frame store 34.

Thedecoder 10 outputs the decoded video frames stored in thereference frame store 34 to be displayed on a display device. Because of the downsampling and pre-parsing operations, the decoded video frames are of a lower resolution than the resolution at which the frames were originally encoded.

The known reduced resolution video decoder illustrated in FIG. 2 can be implemented at lower cost than the standard full resolution decoder illustrated in FIG. 1. However, the knowndecoder 10 has the disadvantage that the images are of reduced resolution due to the use of downsampling.

Not only is it important that modern video decoders be capable of being implemented relatively efficiently, but is also likely to be important that they be capable of decoding video data in such a manner that video sequences, e.g., movies, can be displayed at height to width ratios other than that at which they were encoded. Such capability is important because, e.g., it allows images encoded at height to width ratios found in movie screens or high definition television (HDTV) to be played back on screens with different, e.g., standard definition television (SDTV) height to width ratios.

Currently, two techniques are commonly used for displaying images which were filmed or encoded at one height to width ratio on a screen that supports a second height to width ratio. The first technique is what is commonly referred to as "letter boxing". In this technique, the entire image is displayed at the original height to width ratio with black borders being added to, e.g., the top and bottom of the screen. By adding borders in this manner the portion of the screen which is not being used to display the image is blacked out.

The second technique for displaying images at a height to width ratio which is different from the images' original height to width ratio is commonly known as pan-and-scan. Using this technique, each image's center of interest is identified and the portion of the image, corresponding to the available screen dimensions, immediately surrounding the center of interest is selected for display. The portions of the image which are cropped are frequently referred to as side panels because one such panel is normally located on each side of the center of interest. The center of interest may be the center of the image but in many cases, e.g., in action scenes, may be off center. The center of interest in a sequence of images can change in position from one image to the next.

Notably, the MPEG specification allows for, but does not mandate, the transmission of center of interest information as part of an encoded bitstream.

While the known methods of reducing the cost of video decoders, such as the use of downsampling, have helped lower the cost of such devices, there still remains a need for new and improved methods and apparatus for reducing the cost of video decoders.

It is desirable that at least some of the cost saving methods and apparatus be suitable for use with full resolution video decoders since, in some applications reductions in resolution of the type caused by downsampling are unacceptable. It is also desirable that at least some cost reduction methods and apparatus be capable of being used in video decoders which support various display modes including, e.g., pan-and-scan.

SUMMARY OF THE PRESENT INVENTION

As discussed above, the present invention is directed to methods and apparatus for reducing the cost of video decoders, e.g., MPEG and ATSC compatible video decoders.

Various methods and apparatus of the present invention are directed to decoding frames as a function of whether all or a portion of an image is to be viewed. For example, in one embodiment, when the decoder is operated in a pan-and-scan mode of operation encoded data corresponding to B-frame side panels which will not be viewed is discarded. By discarding B-frame side panel data, the amount of processing required to decode a B-frame is reduced as well as the amount of memory required to store decoded B-frames prior to display. In the case of I and P frames to be decoded and displayed in pan-and scan mode, in various embodiments, side panel portions of the I and P frames are decoded using methods which reduce processing and/or storage requirements as compared to those associated with decoding the displayed image portions.

In one particular embodiment, side panel information is decoded and stored, e.g., through the use of downsampling, at a lower resolution than the image portions that are displayed. In another embodiment, P frame side panels are decoded using simplified inverse quantization techniques while the image areas to be displayed are decoded using standard (MPEG compliant) inverse quantization techniques.

Using decoding techniques of the present invention, processing and storage efficiencies can be achieved in pan-and-scan mode operation while still maintaining a representation of an entire I or P frame in the event that side panel data is required for use in a subsequent prediction.

In order to reduce decoding hardware and processing requirements methods and apparatus of the present invention are directed to improved techniques for performing inverse quantization operations on, e.g., predictively coded data. The inverse quantization methods and apparatus of the present invention are well suited for use in both full and reduced resolution video decoders. In various embodiments of the present invention, as will be discussed below, the simplified inverse quantization method of the present invention is used when decoding, e.g., B-frames, but not I or P frames. In other embodiments, the inverse quantization techniques of the present invention are used with both P and B frames.

The simplified inverse quantization techniques of the present invention reduce the amount of processing required to perform an inverse quantization operation on predictively coded data by treating an additive factor k used when performing inverse quantization arithmetic operations as always being zero and/or by eliminating the mismatch control operation performed as part of a standard inverse quantization operation.

Various other methods and apparatus of the present invention, suitable for use with decoders, e.g., various reduced resolution video decoders, which perform drift reduction filtering, are also described herein. Such methods and apparatus involve dynamically varying the complexity of performed drift reduction filtering as a function of whether luminance or chrominance data is being decoded. Dynamic control of drift reduction filtering complexity is particularly well suited for use in decoders, e.g., reduced resolution video decoders, which normally perform drift reduction filtering.

The methods and apparatus of the present invention can be used in a wide variety of applications. For example, they can be used to implement television receivers, video recorders, computers and a host of other devices which decode video data. Many of the methods of the present invention may be implemented in software executed on, e.g., a general purpose computer, or by dedicated hardware, e.g., integrated circuits used to implement a video decoder.

Numerous additional features and embodiments of the present invention are discussed in detail in the description which follows.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 illustrates a known full resolution video decoder.

FIG. 2 illustrates a known downsampling video decoder.

FIG. 3 illustrates a first video decoder implemented in accordance with an exemplary embodiment of the present invention.

FIG. 4 illustrates a video decoder, implemented in accordance with the present invention, capable of decoding various video data either at full resolution, reduced resolution, or with different portions at different resolutions.

FIG. 5 illustrates a known circuit for performing an inverse quantization operation.

FIG. 6 illustrates an inverse quantization circuit implemented in accordance with the present invention.

DETAILED DESCRIPTION

As discussed above, the present invention is directed to methods and apparatus for reducing the cost of video decoders.

A full resolution, i.e., non-downsampling,video decoder 300 implemented in accordance with one embodiment of the present invention will now be discussed with reference to FIG. 3.

Elements in FIG. 3 which bear the same names as elements of the known decoders 100, 200 are similar to the like named elements but may include slightly different functionality in accordance with the present invention, as will be discussed below.

As illustrated in FIG. 3, thevideo decoder 300 comprises a syntax parser andVLD circuit 304, aninverse quantization circuit 306, aninverse DCT circuit 318, asummer 322,switch 324,memory 330, pair of motion compensatedprediction modules 316 and 317 and a select/average predictions circuit 328. In addition, thedecoder circuit 300 includes a sidepanel ID module 340, an inverse quantization (IQ)control module 342 and a B-framememory management module 344. As illustrated adisplay device 350 may be coupled to the output of the decoder'smemory 330.

Thememory 330 includes a codeddata buffer 332, an I andP frame store 334 and a B-frame store 335. The various components of thedecoder circuit 300 are coupled together as illustrated in FIG. 3. The B-frame store 335 is segmented from the I &P frame store 334. This is because, in modes of operation in which B frame side panels are not stored in accordance with the present invention, less memory is used to store each decoded B frame than is used to stored each decoded I or P frame.

Variable length encoded data in the form of a bitstream and display mode control information are both input to thedecoder 300. The variable length encoded data may be, e.g., an MPEG compliant bitstream representing a sequence of images in the form of I, P and B frames. Information on each encoded image's center of interest may be included in the encoded bitstream.

Received encoded video data is stored in a codeddata buffer portion 332, of thememory 300, prior to being supplied to the syntax parser andVLD circuit 304.

The display mode control information may be obtained, e.g., from a user controlled switch coupled to thedecoder 300. This information indicates, e.g., whether complete images included in the variable length encoded bitstream are to be displayed or only a portion thereof. The display mode control information may indicate, e.g., that a user has selected to view the received images in a pan and scan format. During pan and scan mode operation entire received images are not to be displayed but rather only as much of the image that can fit on the display device's screen given its height to width ratio.

The display mode control information is supplied to, and used by, both the sidepanel ID module 340 and B-framememory management module 344.

In addition to receiving the display mode control information, the sidepanel ID module 340 receives, from the syntax parser andVLD circuit 304, information about the center of interest of each image being decoded when such information is available from the encoded bitstream. In one embodiment, when center of interest information is not available, the center of interest is assumed to be the center of each received frame or image.

The sidepanel ID module 340 determines from the received display mode control information the display mode of operation being used. It also determines, as a function of received center of interest information, which portions of an image lie outside of the primary region of visual interest. In the case of pan-and-scan mode operation this involves determining what portions, if any, of the received encoded images will not be displayed. Portions of the image which will not be displayed are considered to be outside the primary region of visual interest.

Thus, in pan-and-scan mode operation, the sidepanel ID module 340 identifies the image side panels which will be cropped prior to display. This side panel information is provided to theIQ control module 342 and syntax parser andVLD circuit 304.

The syntax parser andVLD circuit 304 is responsible for parsing and variable length decoding the received encoded bitstream. As part of this process, data, e.g., macroblock data, corresponding to individual I, P and B frames is identified.

It is possible, although unlikely, for I and P frame image data outside the displayed image area, e.g., image side panel data, to be used as anchor frame data for subsequent predictions. However, since B-frames are not used as anchor frames, B-frame data, including side panel data, need not be stored for use in subsequent predictions. As part of the decoding process, the B-frames are normally stored prior to display to allow output of the decoded frames to the display device at the desired time intervals. In particular, when the encoded bitstream represents interlaced video with frame structured pictures, or progressive video at a lower frame rate than the display rate, B frames are generally stored in memory to allow the time of decoding to be decoupled from the time of display.

Recognizing that B-frame data may have to be stored prior to display but not for use as anchor frame data, in one embodiment of the present invention, the syntax parser andVLD circuit 304 discards B-frame data corresponding to image areas, e.g., side panels, which will not be displayed. This eliminates the need for the remaining circuits of thedecoder 300 to process and store data corresponding to portions of B-frames which will not be displayed. Thus, by operating the syntax parser andVLD circuit 304 to discard received data corresponding to image portions which will not be displayed or used for predictions in accordance with the present invention, a reduction in the amount of processing required to decode an encoded bitstream is achieved.

In the above described manner, when the displayed picture area is smaller than the area of the picture or image being decoded, computational savings are obtained by controlling thedecoder 300 to skip over bits corresponding to B-frame macroblocks that lie outside the displayable image area. This control function, performed by the sidepanel ID module 340, saves memory and computations, without affecting the delivered picture quality, since B frames are not used for decoding subsequent pictures.

An information data output of the syntax parser andVLD circuit 304 is coupled to the input of the motion compensatedprediction modules 316, 317, and to a second input of theIQ control module 342. The information supplied to the motion compensatedprediction module 306 andIQ control module 342 by thesyntax parser 304 includes, e.g., frame and macroblock type information.

TheIQ control module 342 is used to control theinverse quantization circuit 306. Theinverse quantization circuit 306 is capable of performing a standard inverse quantization operation in accordance with the MPEG standard, or a simplified inverse quantization operation, e.g., a non-exact inverse quantization operation in accordance with the present invention. The IQ control module controls the type of inverse quantization operation that is performed. In one embodiment, it does this as a function of the frame type and the position of the data being decoded relative to an image's center of interest.

In one particular embodiment, theIQ control module 342, controls theIQ circuit 306 so that a standard inverse quantization operation is performed on intra-coded frames, an exact inverse quantization operation is performed on the portion of P-frames which is to be displayed, a simplified inverse quantization operation is performed on the data corresponding to the side panels of P-frames; and a simplified inverse quantization operation is performed on B-frame data output by the syntax parser andVLD circuit 304.

Operation of theinverse quantization circuit 306 which uses the new and novel simplified inverse quantization methods of the present invention to processes predictively coded data will be discussed in detail below. A data output of theinverse quantization circuit 306 is coupled to an input of the inverse discrete cosine transform (IDCT)circuit 318.

TheIDCT circuit 318,switch 324,summer 322, motion compensatedprediction modules 316, 317 and select/average predictions circuit 328 are all coupled together as illustrated in FIG. 3. The arrangement of thesecomponents 316, 317, 328 is the same as or similar to the way like named components of the known decoder of FIG. 1 are coupled together. While thesecomponents 318, 322, 324, 306, and 328 serve the same or similar functions as the like named elements of the known decoder 100, they are capable of processing B frame data representing an image of the same or a different size than the processed I and P frames. Accordingly, thesecomponents 318, 322, 324, 316, 317 and 328 are capable of processing the B frame data even when B-frame side panel data has been discarded by the syntax parser andVLD circuit 304.

The output of theswitch 324, via which decoded video data is supplied to thememory 330 for storage prior to display and/or use as reference frame data, is coupled to the B-framememory management module 344 and to the I andP frame store 334. The decoded video data representing I and P frames is stored in the I &P frame store 334. B-frame data is stored in the B-frame store 335, prior to display, via the B-framememory management module 344.

In addition to decoded B-frame data, the B-framememory management module 344 receives as an input, the display mode control information. The B-framememory management module 344 is responsible for controlling the locations at which the decoded data representing B-frames is stored in thememory 330 to insure the efficient utilization of available memory. In some modes of operation entire B-frames are stored in the B-frame store 335. In other modes of operation, e.g., pan-and-scan modes of operation, in accordance with the present invention, only the portions of the decoded B frames which are going to be displayed are stored in the B-frame store 335. The B-frame memory management module uses the display mode control information to determine what B-frame data is to be stored and the memory addresses at which B-frame data should be stored. In this manner efficient use of the available memory is achieved.

In one particular embodiment, the B-frame management module is sensitive to display mode control information, and acts to allocate memory for temporary storage of the displayable portion of B-frames, when pan-and-scan display mode is selected. Since decoded image data corresponding to side panel B frame macroblocks are not stored or displayed in this mode of operation a reduced amount of memory is allocated by the B-framememory management module 344 for the storage of B frames when pan-and-scan mode is selected.

As discussed above, one feature of the present invention is directed to performing a simplified inverse quantization operation on at least some of the predictively coded data supplied to theinverse quantization circuit 306.

The inverse quantization process is used extensively in the digital processing of signals such as digitally compressed images and digitally compressed video.

Quantization is the process that is used in the digital processing of signals in which an element from a finite set of digital codewords is used to represent, approximately, the value of a continuous-amplitude sampled signal. The digital codewords that are produced by the quantization process for an input sample represent an approximation of the original amplitudes of the signal being processed.

The opposite process of quantization is appropriately called inverse quantization. The inverse quantization process takes as its input a digital codeword from a finite set of codewords and produces a so called reconstruction level that is an approximation of the original amplitude of the sample.

The international standard ISO/IEC 13818-2 which describes MPEG-2 defines methods for the inverse quantization of DCT coefficients. The ATSC A/53 standard, by reference to MPEG, defines the inverse quantization methods to be the same as those of the MPEG standard. In both of these standards, the inverse quantization of one 8×8 block of coefficients, with the exception of the DC coefficients of intra-coded blocks, is specified by equations (1)-(5) which are set forth below. A known inverse quantization circuit for implementing a standard inverse quantization operation in accordance with equations (1)-(5) is illustrated in FIG. 5.

As illustrated, the knowncircuit 500 includes an Inverse QuantizationArithmetic circuit 502, asaturation circuit 504 and amismatch control circuit 506.

The inputs to theinverse quantization circuit 502 are QF[v][u], W[w][v][u] and a quantizer_-- scale

where:

QF[v][u] is a two dimensional array of digital codewords or quantized DCT coefficients, W[w][v][u] is a quantizer matrix, and quatizer_-- scale is a common scaling factor used for one or more macroblocks. The parameters v and u are used to index each DCT coefficient and the parameter w depends upon the coding type (INTRA or NON-INTRA and the color component (luminance or chrominance.) In the decoder 100, these values are part of the data provided by the syntax parser andVLD circuit 14 to theinverse quantization circuit 16.

The known inverse quantizationarithmetic circuit 502 implements equations (1)-(3) set forth below. ##EQU1##

The coefficients F"[v][u] resulting from the processing performed by the inverse quantizationarithmetic circuit 502, are processed by a saturation stage to ensure that the reconstructed values lie within an allowed range. The saturation operation, which yields the coefficients F'[v][u] performed by thesaturation circuit 504 is set forth as equation (4) below. ##EQU2##

The final step in the standard inverse quantization process is to perform a mismatch control operation to yield the final reconstructed DCT coefficients F[v][u] which may then be supplied to theIDCT circuit 18. Themismatch control circuit 506 is responsible for performing the mismatch control operation set forth below as equation (5) using the coefficient values F'[v][u] output by thesaturation circuit 504. ##EQU3##

Equations (1)-(5) are specified in the international standard ISO/IEC 13818-2 standard. For the inverse quantization operation of quantized DCT coded data to be compliant with the standard, equations (1)-(5) must be used to implement the inverse quantization process for the coded blocks.

Table I below shows an analysis of the number of discrete operations that are required to implement the equations assuming the most costly case, in terms of the number of required computations, of performing an inverse quantization operation on NON-INTRA coded data.

              TABLE I                                                     ______________________________________                                    Number of operations required to implement equations (1)-(5)              operations per coefficient Total for 64                                   Equation                                                                          Add/Sub  Mult/Div Compare Total                                                                          coefficients                       ______________________________________                                    (1)     1        4        0       5    320                                 (2),   0        0        2       2    128                                (3)                                                                       (4)     0        0        2       2    128                                (5)     65*      0         2*     --    67                                                Total operations                                                                     643                                            ______________________________________                                     *for all 64 coefficients

An accepted method of obtaining efficient implementations of algorithms is to express the algorithm as a linear operator and then find some sparse factorization of the linear operator. While this has worked well as a technique for simplifying IDCT operations it is not applicable to the problem of computing the inverse quantization of NON-INTRA coded data because it is not possible to represent exactly, the equations (1)-(5) as a linear operator.

Recognizing that it is not possible to use a linear operator formulation to represent exactly, equations (1)-(5) in a simplified form, the inventors of the present application set about finding other ways to simplify the inverse quantization process. The inventors of the present application realized that it is possible to approximate the result of the processing achieved by performing equations (1)-(5) without having to perform all the processing required to fully implement all of these equations. Accordingly, various embodiments of the inverse quantization method of the present invention, approximates the overall processing results that would normally be achieved by accurately implementing equations (1)-(5).

In particular, various embodiments of the present invention focus on simplifying the inverse quantization process by eliminating equation (2) and treating the additive factor k found in equation (1) as being 0 when processing both INTRA and NON-INTRA coded data. Other embodiments of the present invention are directed to eliminating the use of the mismatch control circuit and thereby eliminating the mismatch control operation set forth in equation (5) from the inverse quantization process. In accordance with the present invention these inverse quantization simplification techniques may be selectively applied, as discussed below. The application of the inverse quantization techniques, e.g., on a macroblock or block basis, can depend on the type of frame or image being processed and/or the location of the portion of the image represented by the set of quantized DCT coefficients being processed.

The inverse quantization methods of the present invention may result in a less than exact reconstruction of the image data being processed. However, in most cases the inverse quantization techniques of the present invention produce satisfactory results in terms of image quality while providing advantages in terms of speed and implementation costs. That is, the inverse quantization methods of the present invention tend to provide for a fast and hardware efficient inverse quantization process.

The inverse quantization methods of the present invention make it is possible to reduce the computational cost of the inverse quantization process while maintaining a high level of picture quality. As discussed above, the simplified non-exact inverse quantization processes of the present invention used to process P and B frames, differ from the standard, i.e., MPEG and ATSC specified inverse quantization operations, in the application of the additive factor k present in equation (1) and the use of mismatch control as set forth in equation (5). TheIQ control module 342 is responsible for determining when the simplification involving treating the additive factor k as being 0 for both NON-INTRA coded and INTRA coded data is employed and when performing a mismatch control operation is to be skipped.

FIG. 6 illustrates aninverse quantization circuit 600 implemented in accordance with one embodiment of the present invention. Thecircuit 600 includes an inverse quantizationarithmetic circuit 602, asaturation circuit 504, and amismatch control circuit 606 coupled together in series as illustrated in FIG. 6. Theinverse quantization circuit 600 also includes a multiplexer (MUX) 608. Themultiplexer 608 receives, as first and second data inputs, the coefficients output by themismatch control circuit 606 and the coefficients output by thesaturation circuit 504, respectively.

In the FIG. 6 embodiment, a two bit IQ control signal output by, e.g., theIQ control circuit 342, is supplied to control inputs of the inverse quantizationarithmetic circuit 602 andMUX 608. The first bit of the IQ control signal is used to control when the inverse quantizationarithmetic circuit 602 is to treat the additive factor k as being 0 for both NON-INTRA and INTRA coded data. The second bit of the control signal is used to control which input of theMUX 608 is to be output, e.g., to an IDCT circuit, as the DCT coefficients resulting from the inverse quantization operation performed by thecircuit 600.

By asserting the bit used to control theMUX 608, the inversequantization control module 342 can cause the signal supplied to the second input of theMUX 608 to be output as the result of the inverse quantization operation. This eliminates the use of themismatch control circuit 606 when performing an inverse quantization operation.

Various embodiments of the present invention which relate to selectively using 0 as the value of the additive factor k and/or eliminating the mismatch control operation from the inverse quantization process will now be discussed.

One embodiment of the present invention involves the use a non-exact (e.g., non-MPEG-2 compliant) inverse quantization process for predictively coded (P and B) frames while using an MPEG-2 compliant inverse quantization process for (I) frames.

Normally, the additive factor k, is determined for each quantized coefficient of a set of such coefficients Q[v][u] and thus, the value of k can normally vary as a function of the value of each non-intra-coded coefficient of the set of quantized coefficients Q[v][u].

In one embodiment of the present invention, the value of k is set to zero for all NON-INTRA macroblocks when performing an inverse quantization operation on P and B frame data. In such an embodiment, the value of k is treated as being 0 for all elements of a block of coefficients representing a portion of a predictively coded frame, e.g., P or B frame.

In another simplified inverse quantization embodiment, application of a non-zero value as the additive factor k is restricted to only NON-INTRA macroblocks in P frames. This implementation is described in Eq. (6) below. ##EQU4##

In this case I and P frames will be reconstructed in a fully MPEG compliant manner, but B frames will be reconstructed using the simpler inverse quantization process of the present invention. Since B frames generally require more processing resources than I or P frames to perform an inverse quantization operation the simplified inverse quantization process of the present invention can provide significant hardware savings even when applied only to B frames. Furthermore, by limiting application of the simplified inverse quantization technique of the present invention to B-frames, propagation of errors introduced as a result of the simplified inverse quantization process, into additional frames, is avoided. This is because unlike P-frames, B-frames are not used as anchor frames.

The effect of treating the additive factor, k, as being 0 in accordance with the present invention is to impart, to the images decoded using this technique, a degree of quantization noise that is above what is normal. Visually this effect can be characterized as increased background noise. For the quality loss due to the used of the simplified inverse quantization process to be noticed, the level of noise present in the image must be relatively high.

Significantly, when the simplified inverse quantization techniques of the present invention are applied to frames which are part of a Group of Pictures ("GOP"), the GOP structure causes a refreshing of the picture at every instance of an I or P-frame. This fact prevents the quantization errors in P-frames from accumulating indefinitely.

In the case where application of the simplified quantization techniques of the present invention are limited to B-frames as in equation (6), the additional quantization noise, resulting from the methods of the present invention, is restricted to each individual B-frame, and no accumulation of quantization noise occurs.

In accordance with another embodiment of the present invention mismatch control is not carried out for macroblocks being processed using the simplified non-compliant, inverse quantization process of the present invention. Mismatch control is intended to prevent undesirable effects when an MPEG or ATSC standard inverse quantization operation is performed. It has been discovered that the mismatch control operation performed during a standard inverse quantization operation provides little or no benefit when employing the simplified non-standard inverse quantization process of the present invention.

Thus, in accordance with the present invention further reduction in the computational cost of performing an inverse quantization operation is achieved by not performing a mismatch control operation.

By, e.g., setting k=0, in accordance with the present invention when processing I, P and B frame data as set forth in equation (7), equation (1) is reduced to equation (8).

k=0                                                        (7) ##EQU5##

Equation (8) has the advantage that it can be implemented using one multiplication per coefficient being processed.

By elimination of the mismatch control operation as discussed above, further computation savings are achieved. Table II below shows the reduction in computational cost that may be obtained using the non-exact inverse quantization process with k=0 and the mismatch control operation being eliminated.

              TABLE II                                                    ______________________________________                                    Number of operations required for simplified inverse                      quantization in accordance with Equation (8) and                          eliminating the mismatch control operation                                operations per coefficient Total for 64                                   Equation                                                                          Add/Sub  Mult/Div Compare Total                                                                          coefficients                       ______________________________________                                    (8)     0        1        0       1     64                                (4)     0        0        2       2    128                                                Total operations                                                                     196                                            ______________________________________

The results set forth in Table II is a 70% reduction in the number of operations as compared to those used to implement a standard inverse quantization operation set forth in Table I.

Various features of the present invention, such as discarding of B-frame data that will not be viewed without decoding it and the performing of a simplified inverse quantization operation on all or some predicted image data have been explained with reference to the full resolution video decoder illustrated in FIG. 3. These features and additional features of the present invention may be used to implement a multi-resolution or reduced resolution video decoder as well.

Various features of the present invention will now be discussed with reference to themulti-format video decoder 400 illustrated in FIG. 4. Thedecoder 400 illustrated in FIG. 4 is capable of decoding, for example, high definition television (HDTV) signals at a reduced resolution, e.g., standard definition television (SDTV) resolution. It is also capable of decoding standard television signals at full resolution. This is accomplished via the selective application of downsampling and/or other data reduction techniques.

FIG. 4 illustrates a multi-format video decoder implemented in accordance with one embodiment of the present invention. Components which perform the same or similar functions to those of previously described components bear the same names and will not be described again in detail. Notably, while many of the components of thedecoder 400 perform similar functions to those of the known decoders, many of the components in the FIG. 4 embodiment differ from those of like named known components in that many of the components of the FIG. 4 decoder are responsive to control signals from both sidepanel ID module 440, which is implemented in accordance with the present invention, and aresolution control module 403.

As illustrated in FIG. 4, thedecoder 400 comprises anoptional preparser 401, amemory 430, syntax parser andVLD circuit 404,inverse quantization circuit 406,IDCT circuit 418,downsampler 420,summer 422,switch 424, first and second motion compensatedprediction modules 416, 417, a select/average predictions circuit 428, a B-framememory management module 444 and adisplay 450. In addition, in accordance with the present invention, thedecoder 400 includes a sidepanel ID module 440,IQ control module 442 andresolution control module 403. The various components of thedecoder 400 are coupled together as illustrated in FIG. 4.

In the FIG. 4 embodiment, thememory 430 is divided into a codeddata buffer segment 432, an I &P frame store 434 and a B-frame store 435. As in the case of the FIG. 3 decoder embodiment, the B-framememory management module 444 receives display mode control information and decoded B-frame data. The B-framememory management module 444 controls the storage of decoded B-frame data so that thememory space 435 is used efficiently, particularly when data representing portions of received B-frames, e.g., B-frame side panels, is discarded prior to storage.

In addition to the components discussed above, thedecoder 400 includes, in accordance with the present invention, a sidepanel ID module 440, anIQ control module 442, and aresolution control module 403.

The sidepanel ID module 440 receives the display mode control information via a first input. In addition, from the syntax parser andVLD circuit 404, it receives information about the center of interest of the received encoded frames when such information is available from the encoded received encoded bitstream. When such center of interest information is not received from the syntax parser andVLD circuit 404, it is presumed to be the center of the image.

The sidepanel ID module 440 identifies, image areas outside the center of interest, e.g., image side panels, and provides this information to theIQ control module 442, syntax parser andVLD circuit 404,inverse quantization circuit 406,JDCT circuit 418,downsampler 420,resolution control module 403, and drift reduction filters 426, 427 of the motion compensatedprediction modules 416, 417, respectively. The various circuits which receive the side panel information use this information in determining the processing to be applied to particular data.

TheIQ control module 442 andresolution control module 403, perform control functions. In particular, theIQ control module 442, determines when a simplified, as opposed to standard MPEG or ATSC compliant, IQ operation should be performed on predicted data. Theresolution control module 403 controls the resolution at which various image portions will be decoded. It can also control the particular type of simplified inverse quantization operation that is performed as will be discussed below.

Theresolution control module 403 receives as its inputs frame, motion vector, and other information from the syntax parser andVLD circuit 404 from which it can determine, e.g., whether an HDTV signal or an SDTV signal is being processed. In addition, the resolution control module receives center of interest and side panel information from the side panel ID module.

In accordance with one embodiment of the present invention, theresolution control module 403 uses the received input signals to determine which images, and/or image portions, are to be decoded at full resolution, e.g., the resolution at which they were encoded, and which images or image portions are to be decoded at one or more reduced levels of resolution.

In one embodiment of the present invention, theresolution control module 403 is responsible for controlling the decoding of HDTV images such that they are downsampled and displayed at a maximum resolution equivalent to SDTV resolution. This allows HDTV images to be decoded using the same or approximately the same amount of memory required to implement an SDTV decoder. In the same embodiment the resolution control module controls thedownsampler 420 so that SDTV images are not downsampled but, instead, decoded and displayed at full resolution.

To further reduce the amount of memory required to implement a video decoder and or the amount of processing that must be performed by a video decoder, various embodiments of the present invention are directed to decoding different portions of the same image or frame at different resolutions and/or processing different image portions using processing techniques which differ in terms of their computational complexity. In various embodiments, the resolution and/or processing used to decode a particular image portion depends on the visual importance of the image portion being decoded and/or whether or not the decoded image portion is to be displayed.

Theresolution control module 403, in combination with the sidepanel ID model 440, is responsible for controlling the resolution and processing of various image portions. In one such embodiment, theresolution control module 403 controls thevarious circuits 406, 418, 420, 416, 417, 428 of thedecoder 400 to: 1) perform less computationally intensive processing operations on image areas which will not be displayed or which are further away from the image's center of interest than other image portions and/or to 2) discard data, not process, and/or to reduce the amount of data used to represent image areas which will not be displayed or which are furthest away from an image's center of interest.

For example, when processing SDTV images to be displayed in pan-and-scan mode operation, theresolution control module 403 controls the downsampler to downsample I and P frame side panel data while outputting the data corresponding to the image area to be displayed without performing downsampling thereon. Notably, theinverse quantization circuit 406 may also be controlled to perform simplified inverse quantization operations on predicted data corresponding to side panel macroblocks while performing standard inverse quantization operations on data corresponding to predicted macroblocks in the area to be displayed.

In at least one embodiment, when processing SDTV images which will be completely displayed, the resolution control module controls thedownsampler 420 to perform downsampling on the image being decoded. For example, the portion of the image at the image's center of interest may not be downsampled at all, portions of the image at a first distance away from the center of interest may be downsampled at a first rate, and portions of the image furthest away from the center of interest may be downsampled at a second rate which is greater than the first downsampling rate.

An image is normally represented using chrominance as well as luminance information. The chrominance and luminance information is frequently represented using separate blocks of data to represent the chrominance and luminance information corresponding to the same portion of an image. Thedownsampler 420 can be used to independently downsample chrominance and luminance data corresponding to the same portion of an image. Theresolution control module 403, can, and in one embodiment is, used to control thedownsampler 420 and thus the resolution at which chrominance and luminance data is stored.

Humans are relatively insensitive to distortions introduced in the chrominance (color) components of an image compared to the luminance components of an image. In one embodiment of the present invention, chrominance and luminance data corresponding to the same portion of an image are stored at different resolutions allowing for a relatively efficient use of available memory and processing power. In such an embodiment, the downsampling of chrominance data is performed, e.g., in the absence of downsampling of corresponding luminance data. Alternatively, chrominance data is downsampled at a greater rate than luminance data corresponding to the same portion of the image as the chrominance data being downsampled.

In accordance with the present invention, the computational complexity of chrominance data processing can be reduced in several ways including using reduced precision arithmetic, using reduced order prediction filters, or processing chrominance blocks at reduced spatial resolution relative to that of the collocated luminance blocks.

In thedecoder 400, the motion compensatedprediction modules 416, 417 each include an adaptivedrift reduction filter 426. Thefilter 426 may be used, e.g., to reduce the amount of drift introduced as a result of downsampling. The motion compensated prediction modules, 416, 417 also include scaling circuits, e.g., interpolation and/or decimation circuits, capable of converting stored image portions to the resolution at which the motion vectors are to be applied. In the case where different portions of a reference frame are stored at different resolutions, the scaling circuits included in theprediction modules 416, 417, in one embodiment, scale image portions stored at different resolutions by differing amounts so as to match that of the macroblock being processed.

As discussed above, one feature of the present invention is directed to varying the complexity of prediction operations, and prediction filtering complexity in particular, as a function of whether luminance or chrominance data is being decoded. In this context, complexity may be interpreted in terms of the number of gates required to implement the mathematical processing operations which must be performed to complete a task such as filtering. Accordingly, a more complex filtering operation will require more mathematical operation's such as additions and/or multiplications, and thus more gates to implement, than a less complex operation.

In accordance with this feature of the present invention the complexity of the prediction filtering is varied based on whether luma or chroma information is being processed. In particular advantage is taken of the fact that the human visual system is relatively insensitive to distortions introduced in the chroma components of a picture. Accordingly, in one embodiment more accurate, but more computationally intensive, prediction filtering is performed by thefilters 426 when processing luminance information and less accurate, but computationally simpler, prediction filtering is performed when processing chrominance information.

Thus, in the FIG. 4 embodiment, the chroma adaptivedrift reduction filter 426 is a filter whose operation depends on whether luminance or chrominance information is being processed. Thefilter 426 operates by employing higher complexity filtering when processing luminance components than when processing chrominance picture components. In this manner, the reconstructed quality of the luminance components will generally be more accurate than that of the chrominance components, and there will be a relative reduction in the computational resources used in computing the chrominance components.

Claims (15)

What is claimed is:

1. A method of performing an inverse quantization operation, the method including the steps of:

determining if a set of quantized coefficients being processed corresponds to an intra-coded frame or a bi-directionally coded frame; and

controlling the value of an additive factor, k, used in performing an inverse quantization arithmetic operation to be zero when processing quantized coefficients corresponding to intra-coded and bi-directionally coded frames.

2. The method of claim 1, wherein the set of quantized coefficients is represented by the symbol QF[v][u], further comprising the step of:

determining if a set of quantized coefficients being processed corresponds to a forward predictively coded frame; and

controlling the value of the additive factor, k, used in performing an inverse quantization arithmetic operation when the set of quantized coefficients, QF[v][u] corresponds to a forward predictively coded frame to be as follows:

k=0 if i) QF[v][u] has a value equal to zero and is from a set of intra-coded coefficients or ii) QF[v][u] is from a set of inter-coded coefficients;

k=+1 if QF[v][u] has a value less than zero and is from a set of inter-coded coefficients; and

k=-1 if QF[v][u] is a set of inter-coded coefficients having a value less than 0.

3. The method of claim 2, where, when it is determined that the set of quantized coefficients, QF[v][u] corresponds to a forward predictively coded frame, the method further comprises the steps of:

performing a saturation operation; and

performing a mismatch control operation.

4. A method of performing an inverse quantization operation, the method including the steps of:

performing an inverse quantization arithmetic operation and saturation operation on a set of quantized coefficients to produce a first set of values; and

if the first set of values corresponds to a bi-directionally coded frame:

i. outputting the first set of values as a set of inverse quantized coefficient values; and

if the set of first set of values corresponds to a forward predictively coded frame:

i. performing a mismatch control operation on the first set of values to generate a second set of values; and

ii. outputting the second set of values as a set of inverse quantized coefficient values.

5. A method of processing a non-zero set of inter-coded data QF[v][u], representing a two dimensional array of quantized coefficients of a P-frame, as a function of a quantizer matrix, W[w][v][u], and a scaling factor, quantizer_-- scale, the method comprising the step of:

utilizing a processing circuit to perform an arithmetic operation on the set of data, QF[v][u], to generate a set of data F"[v][u] in accordance with the following equation: ##EQU6##

6. The method of claim 5, wherein the processing circuit is an inverse quantization arithmetic circuit.

7. The method of claim 5, wherein the processing circuit is a programmable computer, the method further comprising the step of: using a computer program to control the programmable computer to perform the arithmetic operation.

8. The method of claim 5, further comprising the steps of:

performing a saturation operation on the set of data F"[v][u], to generate a set of inverse quantized coefficients F[v][u], the saturation operation including the step of implementing an operation using the set of data F"[v][u] in accordance with the following equation: ##EQU7##

9. The method of claim 5, further comprising the steps of: performing a saturation operation using the set of data F"[v][u], to generate a set of intermediate values F'[v][u], the saturation operation including the step of implementing an operation on the set of data F"[v][u] in accordance with the following equation: ##EQU8## and performing a mismatch control operation on the set of intermediate values F'[v][u] to generate a set of inverse quantized coefficients F[v][u].

10. A method of processing a stream of MPEG-2 complaint video data including quantized transform coefficients corresponding to a plurality of I frames and a plurality of B frames, the method comprising the steps of:

performing a non-MPEG-2 compliant inverse quantization operation on at least some of the quantized transform coefficients corresponding to bi-directionally coded frames; and

performing an MPEG-2 compliant inverse quantization operation on at least some of the quantized transform coefficients corresponding to I frames.

11. The method of claim 10, wherein the stream of MPEG-2 compliant video data further includes quantized transform coefficients corresponding to a plurality of predictively-coded frames, the method further comprising the step of:

performing a non-MPEG-2 compliant inverse quantization operation on at least some of the quantized transform coefficients corresponding to P frames.

12. The method of claim 10, wherein the stream of MPEG-2 compliant video data further includes quantized transform coefficients corresponding to a plurality of predictively-coded frames, the method further comprising the step of:

performing an MPEG-2 compliant inverse quantization operation on at least some of the quantized transform coefficients corresponding to P frames.

13. A method of processing inter-coded video data including non-zero inverse quantized transform coefficients, the method comprising the steps of:

performing an inverse quantization operation on the non-zero quantized transform coefficients, the step of performing an inverse quantization operation including the steps of:

i. performing an inverse quantization arithmetic operation on the non-zero quantized transform coefficients to generate an intermediate set of values; and

ii. performing a saturation operation on the intermediate set of values to produce a set of inverse quantized transform coefficients; and

performing a inverse transform operation on the set of inverse quantized transform coefficients to generate a set of pixel value data.

14. A video data processing apparatus, comprising:

means for performing an inverse quantization arithmetic operation on a set of quantized coefficients;

means for performing a saturation operation coupled to the means for performing an inverse quantization operation;

means for performing a mismatch control operation coupled to the saturation operation performing means; and

means for bypassing the mismatch control operation means when processing video data corresponding to B frames.

15. The apparatus of claim 14, wherein the means for performing an inverse quantization arithmetic operation, means for performing a saturation operation and means for performing a mismatch control operation are implemented using a processing unit and software which controls the operation of the processing unit.

Images